What this release publishes, and why so little
A survey of the field recorded 177 entries, each with a name, a category, a free-text descriptor, a service description, a price, an internal note and a URL. It did not record a public source for each individual claim. That single gap determines everything about this release.
A claim without a source is not a small defect that can be labelled around. If this index published “led most psychedelic public offerings” or “approved by the Oregon Health Authority” with nothing behind it, a reader would have no way to distinguish it from a claim that was checked — and the presence of any unsourced claim degrades every sourced one on the page. So the release withholds pricing, service descriptions, credentials, rankings, acquisitions and promotional characterisations entirely, and publishes only what can stand without a source: that a name appeared in the survey, the category it was filed under, and whether a link answered on a stated date.
The descriptor rule
The one field that required judgement is the short scope descriptor. Some are purely descriptive — “Oregon psilocybin licensing; cannabis and emerging industries” makes no claim beyond stating a practice area. Others carry exactly the material withheld everywhere else, in miniature: certifications, superlatives, transaction values, rankings.
Rather than judge them one at a time, a single mechanical rule is applied. A clause is removed if it contains a currency or percentage figure, a scaled number, a superlative, a ranking reference, or a claim of approval, accreditation, certification, licensure, acquisition or exit.
The rule first shipped at record level: any descriptor containing any such clause was dropped whole. That was wrong in a way worth naming, because it destroyed sound information to suppress unsound information sitting next to it. “OHA/HECC-licensed — 9-month/205-hr, Ashland OR” became a blank; the programme length and the town had nothing wrong with them. Forty-nine of 175 records were emptied that way, and because each printed the same notice explaining why, that notice became the single most repeated sentence on the site — which is its own kind of failure to inform.
The rule is now applied clause by clause. A clause survives if it makes no claim and carries at least two words, a lone fragment being closer to damage than description. 151 records publish a descriptor, 25 of them with part removed and the removal counted on the card. 24 have no clause that survives and say so.
Suppression is the cheap half. A claim can also leave the withheld pile by being checked, and 9 have been: the facilitator-training approvals against Colorado’s published list of approved natural-medicine training programmes, and the two state registry records against the registries themselves. Each such record names the source and the date it was read. Two checks came back against the survey: one training programme recorded as Colorado-approved is not on Colorado’s list, and the Oregon licensee directory recorded here as an authoritative full list is described by Oregon as neither comprehensive nor usable for licence verification. Both now say so on the record. Nothing else in this release has been verified this way, and a claim without such a line has not been checked.
The rule is deliberately blunt and it over-withholds. A descriptor that mentions a genuine state approval is suppressed alongside a descriptor that merely boasts, because the index has no source for either and cannot tell them apart. Where a claim of state approval matters — facilitator training is the obvious case — the state holds that fact and publishes it, and that is where a reader should go.
Deduplication
Names are compared after trimming whitespace and lowercasing, and the first occurrence wins. 177 raw entries reduce to 175 published records. This catches exact duplicate names and nothing else: an organisation recorded twice under two different names, or under a former and a current name, is still present twice. No attempt was made to resolve organisational identity, because doing so is a research task and this is a snapshot.
The link check, and what it establishes
Destination links were checked on , separately from the 20 August 2026 entry review. Each recorded URL was requested over HTTPS. Only links that answered are published as links. Hosts that refuse automated requests but serve people normally are treated as reachable, because a refusal to answer a script is not evidence that a site is gone.
What that check establishes is narrow and worth stating exactly: that a server answered at that hostname over HTTPS on that date. It does not establish that the site belongs to the named organisation, that the organisation still operates, that the page describes the same business it described when the survey ran, or that the domain has not changed hands. A parked domain answers. So does a domain someone else has bought.
15 of 175 entries are shown without a link: 13 whose domain did not resolve or served no working HTTPS site when links were checked on 29 August 2026, one whose certificate had expired at the moment of checking, and one that never had a website recorded. They are listed rather than deleted. The fact this index records is that a name appeared in the field; a working website is a separate fact, and losing one does not erase the other. The unreachable entries also cluster in a way that is itself informative — three of the five records under Capital Markets / VC and four of the twenty under Legal.
The failure an automated check cannot see
An HTTPS request that completes is not the same as a website a person can open. A server whose TLS certificate has expired still answers; a browser then refuses to render the page and shows a full-screen security warning instead. To a script that is success. To a reader it is a wall.
One entry in this index is in exactly that state, found by hand rather than by the automated pass, and its link is withheld with the reason and the timestamp shown on the record itself. Certificate expiry is usually temporary and often resolves within hours, which is why the record says what was observed and when rather than marking the organisation as gone.
What is not checked at all
No licence is verified. No credential is verified. No professional standing, insurance, incident history, financial condition or regulatory status is verified. No organisation was contacted. Nobody was asked whether they wished to be listed, and inclusion was not solicited. Nothing on this site should be read as a finding about any organisation’s quality, safety or lawfulness, and nothing here is professional advice.
What a future claim would require
Detailed claims will return only with claim-level provenance. An evidence record for a single proposition must carry: the exact proposition as it will be published, the primary URL, the publisher, the source type, the publication date, the access date, the jurisdiction it applies in, the reviewer, the verification result, the boundary of what the source supports, and a next-review date. That is a high bar deliberately. It is the bar that would let a reader check a claim without trusting the person who wrote it.
Retired labels
Earlier versions carried tier labels — Tier 1, Tier 2, Tier 3 — and an evidence-status grade derived from them. Both are retired. A tier reads as a ranking whatever the caption says, and no ranking was ever performed; the grades were a record of how much the surveyor had written about an entry, not of how well it had been verified. Publishing them under a different name would have preserved the same false signal.
Dates
Entries were last reviewed . Links were last checked . Neither date is generated from the clock. Both are recorded values, and they change only when the work is done again — which means that if this page still reads 29 August 2026 a year from now, the correct inference is that the work has not been redone, and the index should be treated accordingly.
How to check this yourself
Every claim on this site is reproducible without asking anyone. The record count, the category counts and the link-status counts are all derived from the same published data as the pages themselves, and the composition table on the directory page shows the working. The link status of any entry can be confirmed by requesting the destination and comparing the result against the date stated here. Where the two disagree, the check on this page is the one that is out of date — say so through Corrections.